GPU model | Raven | TU102 |
---|---|---|
Architecture | GCN 5.0 | Turing |
Foundry | GlobalFoundries | TSMC |
Manufacturing process | 14 nm | 12 nm |
Number of transistors | 4,940 million | 18,600 million |
Die size | 210 mm² | 754 mm² |
GPU variant | — | TU102-875-A1 |
Release date | May 10th, 2018 | — |
---|---|---|
Generation | Raven Ridge (Vega) | — |
Production | Active | — |
Interface | IGP | — |
Predecessor | Great Horned Owl | — |
Successor | Picasso | — |
Base clock | 300 MHz | 1065 MHz |
---|---|---|
Boost clock | 1280 MHz | 1395 MHz |
Memory clock | System Shared | 1575 MHz 12.6 Gbps effective |
Memory size | System Shared | 16 GB |
---|---|---|
Memory type | System Shared | GDDR6 |
Memory bus width | System Shared | 384 bit |
Bandwidth | System Dependent | 604.8 GB/s |
Shading units | 384 | 4608 |
---|---|---|
TMUs | 24 | 288 |
ROPs | 8 | 96 |
Compute units | 6 | — |
SM count | — | 72 |
Tensor cores | — | 576 |
RT cores | — | 72 |
Cache L1 | — | 64 KB (per SM) |
Cache L2 | — | 6 MB |
Pixel rate | 10.24 GPixel/s | 133.9 GPixel/s |
---|---|---|
Texture rate | 30.72 GTexel/s | 401.8 GTexel/s |
FP16 (half) performance | 1.966 TFLOPS (2:1) | 25.71 TFLOPS (2:1) |
FP32 (float) performance | 983.0 GFLOPS | 12.86 TFLOPS |
FP64 (double) performance | 61.44 GFLOPS (1:16) | 401.8 GFLOPS (1:32) |
Slot width | IGP | Dual-slot |
---|---|---|
TDP | 15 W | 260 W |
Outputs | No outputs | No outputs |
Length | — | 267 mm 10.5 inches |
Suggested PSU | — | 600 W |
Power connectors | — | 1x 6-pin + 1x 8-pin |
Board number | — | PG150 SKU 215 |
DirectX | 12 (12_1) | 12 Ultimate (12_2) |
---|---|---|
OpenGL | 4.6 | 4.6 |
OpenCL | 2.1 | 3.0 |
Vulkan | 1.2 | 1.2 |
Shader model | 6.4 | 6.6 |
CUDA | — | 7.5 |
Release date | — | Unknown |
---|---|---|
Generation | — | GRID |
Production | — | Active |
Interface | — | PCIe 3.0 x16 |